There's a growing worry that AI will break empirical social science -- that agents can p-hack until they find something that "works."
We think that worry deserves to be taken seriously. Our new paper shows that is true empirically and makes it precise: njw.fish/static/paper...
Interesting AmericasBarometer result from @johnsides.bsky.social in this week's @goodauth.bsky.social newsletter. Strong public willingness to trade features of democracy (free elections, free expression) for material and physical security, with some party differences.
mailchi.mp/goodauthorit...
If you want something fun to read between bouts of tarriff-related-scrolling, I have a suggestion
cauldronllc.substack.com/p/hermetica-...
My new column:
If people want the horrors of ICE to end, we must oust the GOP.
All Democrats must vote for the Dem in the primary who can win the general ... and in the general vote blue no matter who.
No 3rd party voting or staying home. That's deadly.
www.usatoday.com/story/opinio...
Most Americans have 3 months of savings and are able to finance normal expenditures in the absence of a missed paycheck.
America is a rich country - most (not all!) Americans are financially comfortable compared to anywhere else (or Americans at previous points in history).
Reading this paper in more detail you actually learn that the effect is found to be null among Democrats and Dem leaners and on partisanship in general and the changes in attitudes they find are concentrated among republicans.
Wrote about this more for Searchlight. It's a huge tragedy and a blow to public opinion research.
open.substack.com/pub/searchli...
This is a huge problem for researchers- there's tons of approval polling, but Gallup's consistency makes it the source of choice for historical comparisons
Gallup says it will stop doing presidential approval polls (after 88 years of doing it) because of "an evolution in how Gallup focuses it's public research". This is eliminating a time series that goes back to FDR.
thehill.com/homenews/med...
Very cool paper- kind of wild that you have to literally pay people to pay attention before this content has an impact
We find that voters are slightly, but not substantially better represented than non-voters. In contrast, people active in multiple forms of nonelectoral participation (like demonstrations, petitions, contacting officials) are better represented.
I kind of love this, although I don't think people who aren't survey nerds would tolerate it. My least favorite thing about polling is the false sense of precision.
(That big heatmap situation is interactive in the actual post)
Anyways I think this is super interesting, check it out
You can compare across filings to see who shares donor populations with who. Everyone pulls from Ossoff's massive donor base, but oddly, Platner and El-Sayed share a bunch as well.
On the blog today: how similar are the donors of major senate candidates? Turns out, something like 60% of them are literally the same people
open.substack.com/pub/cauldron...
we should dramatically expand the house and the senate to improve our n sizes
many different estimates even within their work- the original Bonica et al 2025 found ~1% for the dem moderating, the newer Bonica/Grumbach research note finds anywhere between 0 and .5% depending on the ideology measure, the substack post from them cites +1.4 pp in competitive districts
oh god we're doing it again aren't we
This whole fight comes down to "is 1-2 points a large effect or a small effect" , where the campaign side folks say "that's a large effect" and the academics say "that's a small effect"
It's probably a futile exercise to be trying to explain to the folks in this argument that they largely agree, but still: everyone agrees that incumbents with fundraising advantages do very well. This is why I have so much heartburn over primarying swing district candidates!
Could not disagree more with this, I love my job very much but if I was suddenly not needed for my job, I have a stack of hobbies and side projects waiting to fill that gap.
I want a bright pink shirt that says "life is hard let's take surveys" or something, this will be mega popular among upwards of 5 women in polling
This guy doesn't know that women love to take surveys (or like, presumably he does and is just being a dick)
More stuff from our polling that is also kind of a rorschach test- 23% of Trump voters say ICE tactics are too forceful. Is that a. depressingly low or b. surprisingly high? I lean towards b, personally.
This polling backs up our earlier memo- americans are horrified by ICE and interested in reform, but they still want immigration enforcement to happen. You can consult DFP for polling on how much people don't agree on what "abolish ice" means x.com/DataProgress...
Anyways toplines are linked in the memo!
Also, voters have some specific ideas about how law enforcement should interact with ICE. Checking immigration status on arrest, yes, detaining people at hospitals, absolutely not.
There's been some fascinating different reactions, some people are shocked that the support for changes to ICE is so high, some that it's so low. Polling!
58% of voters want changes that rein in ICE, with the plurality landing on reform (rather than eliminating the agency, or replacing it).
New from Searchlight today, some hopefully helpful polling on ICE. Voters reeaaaally hate ICE detaining Americans and entering homes without warrants.
www.politico.com/news/2026/01...
